Problem statement

1. Loading the data

Visualizing images in dataset

2,3 Exploratory Data Analysis & insights

Finding the mean image for different type of seeds

Contrast Difference

4. Data Preprocessing

Normalizing Data

Guassian blurring

5. Make data compatible

One Hot encoding

Splitting the data into Training testing and validation data

Printing the shape of data

Check train, validation and test sets, if in compatible shapes

We can see from above shape details, that all three are already compatible.

6, 7, 8 Model Building, evaluation and performance improvement

1. CNN model (with 3 conv2D layers)

2. CNN Model with 5 conv2D


Final Loss and accuracy on Validation set

Predict labels for test data

Confusion Matrix

Final Loss and Accuracy

Let's visualize the predicted result

3. CNN without blurring the images

Printing the shape of dataset

Model Summary

Predicting labels for test data

Confusion matrix

Final loss and Accuracy

Let's visualize prediction results¶

9. Conclusion